Text Mining Groupwork 2: Analysis of Twitter Activities by Swiss Universities of Applied Sciences
Fabio Burri, Mischa Haenen, Robin Galeazzi and Timon Galeazzi
2024-05-20
`
Load the tweets and check if they are loaded correctly. We also check the summary for a first interpretation. The summary(tweets) output reveals the following:
- Time Range: Tweets span from 2009 to 2023, with most posted between 2015 and 2020.
- Distribution: The data is unevenly distributed, with the median date in 2018.
- Engagement: Retweets and favorites range from 0 to 267 and 188, respectively, with most tweets having low engagement.
- Universities: The university column indicates tweets from multiple universities, allowing for comparison.
# Set working directory
# getwd()
# setwd("./data/")
# Load data
load("../data/Tweets_all.rda")
# Check that tweets are loaded
head(tweets)## # A tibble: 6 × 14
## created_at id id_str full_text in_reply_to_screen_n…¹
## <dttm> <dbl> <chr> <chr> <chr>
## 1 2023-01-20 17:17:32 1.62e18 1616469988369469… "Im MSc … <NA>
## 2 2023-01-13 07:52:01 1.61e18 1613790954737074… "Was bew… <NA>
## 3 2023-01-12 19:30:01 1.61e18 1613604227141537… "Was uns… <NA>
## 4 2023-01-12 08:23:00 1.61e18 1613436367169634… "Eine di… <NA>
## 5 2023-01-11 14:00:05 1.61e18 1613158809081450… "Wir gra… <NA>
## 6 2023-01-10 17:06:11 1.61e18 1612843252083834… "Unsere … <NA>
## # ℹ abbreviated name: ¹in_reply_to_screen_name
## # ℹ 9 more variables: retweet_count <int>, favorite_count <int>, lang <chr>,
## # university <chr>, tweet_date <dttm>, tweet_minute <dttm>,
## # tweet_hour <dttm>, tweet_month <date>, timeofday_hour <chr>summary(tweets)## created_at id id_str
## Min. :2009-09-29 14:29:47.0 Min. :4.469e+09 Length:19575
## 1st Qu.:2015-01-28 15:07:41.5 1st Qu.:5.604e+17 Class :character
## Median :2018-04-13 13:26:56.0 Median :9.848e+17 Mode :character
## Mean :2017-12-09 15:26:50.7 Mean :9.400e+17
## 3rd Qu.:2020-10-20 10:34:50.0 3rd Qu.:1.318e+18
## Max. :2023-01-26 14:49:31.0 Max. :1.619e+18
## full_text in_reply_to_screen_name retweet_count favorite_count
## Length:19575 Length:19575 Min. : 0.000 Min. : 0.00
## Class :character Class :character 1st Qu.: 0.000 1st Qu.: 0.00
## Mode :character Mode :character Median : 1.000 Median : 0.00
## Mean : 1.289 Mean : 1.37
## 3rd Qu.: 2.000 3rd Qu.: 2.00
## Max. :267.000 Max. :188.00
## lang university tweet_date
## Length:19575 Length:19575 Min. :2009-09-29 00:00:00.00
## Class :character Class :character 1st Qu.:2015-01-28 00:00:00.00
## Mode :character Mode :character Median :2018-04-13 00:00:00.00
## Mean :2017-12-09 02:25:45.00
## 3rd Qu.:2020-10-20 00:00:00.00
## Max. :2023-01-26 00:00:00.00
## tweet_minute tweet_hour
## Min. :2009-09-29 14:29:00.00 Min. :2009-09-29 14:00:00.00
## 1st Qu.:2015-01-28 15:07:00.00 1st Qu.:2015-01-28 14:30:00.00
## Median :2018-04-13 13:26:00.00 Median :2018-04-13 13:00:00.00
## Mean :2017-12-09 15:26:24.68 Mean :2017-12-09 14:59:43.81
## 3rd Qu.:2020-10-20 10:34:30.00 3rd Qu.:2020-10-20 10:00:00.00
## Max. :2023-01-26 14:49:00.00 Max. :2023-01-26 14:00:00.00
## tweet_month timeofday_hour
## Min. :2009-09-01 Length:19575
## 1st Qu.:2015-01-01 Class :character
## Median :2018-04-01 Mode :character
## Mean :2017-11-24
## 3rd Qu.:2020-10-01
## Max. :2023-01-01Start preprocessing the tweets, to calculate the intervalls some additional properties are needed. The preprocessing steps transform raw tweet data into a structured format suitable for analysis. This includes:
- Standardizing Date and Time: Ensures accurate temporal analysis.
- Extracting Features: Derives relevant information (year, month, day, language) for targeted analysis.
- Handling Emojis: Converts emojis to text for sentiment analysis and visualization.
- Data Cleaning: Removes unnecessary tags and standardizes data types.
# Preprocessing Step: Convert date and time to POSIXct and format according to date, year and university. Detect language and extract emojis. The days are sorted from the system locale starting from monday
tweets <- tweets %>%
mutate(
created_at = as.POSIXct(created_at, format = "%Y-%m-%d %H:%M:%S"),
date = as.Date(created_at),
day = lubridate::wday(created_at,
label = TRUE, abbr = FALSE,
week_start = getOption("lubridate.week.start", 1),
locale = Sys.getlocale("LC_TIME")
),
year = year(created_at),
month = floor_date(created_at, "month"),
university = as.character(university),
lang = detect_language(full_text),
full_text_emojis = replace_emoji(full_text, emoji_dt = lexicon::hash_emojis)
)
# Remove Emoji Tags helper funciton
# replace emoji places the emojis in the text as tags and their name, we remove them here
remove_emoji_tags <- function(text) {
str_remove_all(text, "<[a-z0-9]{2}>")
}
# Remove Emoji Tags
tweets$full_text_emojis <- sapply(tweets$full_text_emojis, remove_emoji_tags)
# Store emojis in a sep arate column to analyze later
tweets$emoji_unicode <- tweets %>%
emoji_extract_nest(full_text) %>%
select(.emoji_unicode)Question 1: How many tweets are being posted by the various Universities when? Are there any ‘release’ strategies visible?
Most Active Hours:
Each university has a distinct peak hour for tweeting, often aligning with typical working hours (9 AM - 5 PM). This suggests a strategic approach to reach their target audience when they are most likely online. The most active hours for each university are as follows:
- FHNW: 9 AM
- FH Graubünden: 11 AM
- ZHAW: 5 PM
- BFH: 8 AM
- HES-SO: 10 AM
- HSLU: 9 AM
- OST-FH: 8 AM
- SUPSI-CH: 11 AM
These times typically align with standard working hours, indicating a strategic approach to reach their audience during times they are most likely to be online. It appears that a typical worker is more productive and active on Twitter in the morning, with motivation waning around midday and continuing to decline until the end of the workday.
Most Active Days:
There isn’t a consistent “most active day” across universities. Some favor weekdays, while others show higher activity on weekends. This could reflect differences in their target audience or the nature of their content.
- FHNW: Tuesday
- FH Graubünden: Tuesday
- ZHAW: Wednesday
- BFH: Tuesday
- HES-SO: Tuesday
- HSLU: Thursday
- OST-FH: Friday
- SUPSI-CH: Friday
The pattern also suggests that tweet activity tends to be higher earlier in the week, with motivation and tweet frequency potentially falling as the week progresses.
Release Strategies:
strategy and perhaps a more reactive approach to current events or trends. While universities have peak hours and days, the intervals between tweets vary significantly, indicating a more reactive strategy rather than a rigid release schedule. This variability suggests that the universities might be responding to real-time events or trends rather than sticking to a strict posting schedule.
# Count each tweet by university and hour of the day
tweet_counts_by_hour_of_day <- tweets %>%
group_by(university, timeofday_hour) %>%
count() %>%
arrange(university, timeofday_hour)
# Plot the number of tweets by university and hour of the day
ggplot(
tweet_counts_by_hour_of_day,
aes(
x = timeofday_hour, y = n,
color = university, group = university
)
) +
geom_line() +
facet_wrap(~university) +
labs(
title = "Number of tweets by university and hour",
x = "Hour of day",
y = "Number of tweets"
)
# Show most active hours for each university
hours_with_most_tweets_by_uni <- tweet_counts_by_hour_of_day %>%
group_by(university, timeofday_hour) %>%
summarize(total_tweets = sum(n)) %>%
group_by(university) %>%
slice_max(n = 1, order_by = total_tweets)
print(hours_with_most_tweets_by_uni)## # A tibble: 8 × 3
## # Groups: university [8]
## university timeofday_hour total_tweets
## <chr> <chr> <int>
## 1 FHNW 09 344
## 2 FH_Graubuenden 11 493
## 3 ZHAW 17 580
## 4 bfh 08 497
## 5 hes_so 10 315
## 6 hslu 09 380
## 7 ost_fh 08 44
## 8 supsi_ch 11 330# Show most active hour overall
hour_with_most_tweets <- tweet_counts_by_hour_of_day %>%
group_by(timeofday_hour) %>%
summarize(total_tweets = sum(n)) %>%
arrange(desc(total_tweets)) %>%
slice_max(n = 1, order_by = total_tweets)
print(hour_with_most_tweets)## # A tibble: 1 × 2
## timeofday_hour total_tweets
## <chr> <int>
## 1 11 2356# Count each tweet by university and weekday
tweet_counts_by_week_day <- tweets %>%
group_by(university, day) %>%
count() %>%
arrange(university, day)
# Plot the number of tweets by university and day of the week
ggplot(
tweet_counts_by_week_day,
aes(
x = day, y = n,
color = university,
group = university
)
) +
geom_line() +
facet_wrap(~university) +
labs(
title = "Number of tweets by university and day of the week",
x = "Day of the week",
y = "Number of tweets"
)
# Show most active days for each university
days_with_most_tweets_by_uni <- tweet_counts_by_week_day %>%
group_by(university, day) %>%
summarize(total_tweets = sum(n)) %>%
group_by(university) %>%
slice_max(n = 1, order_by = total_tweets)
print(days_with_most_tweets_by_uni)## # A tibble: 8 × 3
## # Groups: university [8]
## university day total_tweets
## <chr> <ord> <int>
## 1 FHNW Tuesday 575
## 2 FH_Graubuenden Tuesday 751
## 3 ZHAW Wednesday 636
## 4 bfh Tuesday 651
## 5 hes_so Tuesday 415
## 6 hslu Thursday 603
## 7 ost_fh Friday 65
## 8 supsi_ch Friday 461# Calculate time intervals between tweets
find_mode <- function(x) {
ux <- unique(x)
ux[which.max(tabulate(match(x, ux)))]
}
tweets <- tweets %>%
arrange(university, created_at) %>%
group_by(university) %>%
mutate(time_interval = as.numeric(
difftime(created_at, lag(created_at), units = "mins")
))
# Descriptive statistics of time intervals
summary(tweets$time_interval)## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## 0.0 148.2 1128.8 2097.6 2428.3 220707.0 8# setwd("../4.Text-Mining-Groupwork/plots")
unique_years <- tweets$year %>% unique()
# Pilot distribution of time intervals between tweets for each year
for (curr_year in unique_years) {
# Filter data for the specific year
filtered_data <- tweets %>%
filter(year(created_at) == curr_year)
print(ggplot(filtered_data, aes(x = time_interval)) +
geom_histogram(fill = "lightblue") +
facet_wrap(~university) +
labs(
title = paste0(
"Distribution of time intervals between tweets - ", curr_year
),
x = "Time interval (minutes)",
y = "Tweet count"
))
universities <- filtered_data$university %>% unique()
for (uni in universities) {
# Filter data for the specific university
uni_filtered_data <- filtered_data %>%
filter(university == uni)
print(ggplot(uni_filtered_data, aes(x = time_interval)) +
geom_histogram(fill = "lightblue") +
labs(
title = paste0(
"Distribution of time intervals between tweets for ", uni,
" in ", curr_year
),
x = "Time interval (minutes)",
y = "Tweet count"
))
# Calculate mode (most common interval) in hours
most_common_interval_minutes <- find_mode(uni_filtered_data$time_interval)
most_common_interval_hours <- most_common_interval_minutes / 60
print(paste0(
"Most common time interval for ", uni,
" in ",
curr_year,
" is ", most_common_interval_minutes,
" minutes (", most_common_interval_hours, " hours)"
))
}
}
## [1] "Most common time interval for FHNW in 2011 is NA minutes (NA hours)"
## [1] "Most common time interval for FH_Graubuenden in 2011 is 23210.3 minutes (386.838333333333 hours)"
## [1] "Most common time interval for hes_so in 2011 is 1.55 minutes (0.0258333333333333 hours)"
## [1] "Most common time interval for FHNW in 2012 is 17324.65 minutes (288.744166666667 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2012 is 0.9 minutes (0.015 hours)"
## [1] "Most common time interval for ZHAW in 2012 is NA minutes (NA hours)"
## [1] "Most common time interval for bfh in 2012 is NA minutes (NA hours)"
## [1] "Most common time interval for hes_so in 2012 is 22086.35 minutes (368.105833333333 hours)"
## [1] "Most common time interval for FHNW in 2013 is 1.26666666666667 minutes (0.0211111111111111 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2013 is 21879.45 minutes (364.6575 hours)"
## [1] "Most common time interval for ZHAW in 2013 is 0.583333333333333 minutes (0.00972222222222222 hours)"
## [1] "Most common time interval for bfh in 2013 is 65.0833333333333 minutes (1.08472222222222 hours)"
## [1] "Most common time interval for hes_so in 2013 is 36252.5833333333 minutes (604.209722222222 hours)"
## [1] "Most common time interval for supsi_ch in 2013 is 0.783333333333333 minutes (0.0130555555555556 hours)"
## [1] "Most common time interval for FHNW in 2014 is 4.58333333333333 minutes (0.0763888888888889 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2014 is 0.183333333333333 minutes (0.00305555555555556 hours)"
## [1] "Most common time interval for ZHAW in 2014 is 0.05 minutes (0.000833333333333333 hours)"
## [1] "Most common time interval for bfh in 2014 is 153.35 minutes (2.55583333333333 hours)"
## [1] "Most common time interval for hes_so in 2014 is 21986.6 minutes (366.443333333333 hours)"
## [1] "Most common time interval for supsi_ch in 2014 is 37496.4833333333 minutes (624.941388888889 hours)"
## [1] "Most common time interval for FHNW in 2015 is 48918.3 minutes (815.305 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2015 is 1139.9 minutes (18.9983333333333 hours)"
## [1] "Most common time interval for ZHAW in 2015 is 0.316666666666667 minutes (0.00527777777777778 hours)"
## [1] "Most common time interval for bfh in 2015 is 20272.0333333333 minutes (337.867222222222 hours)"
## [1] "Most common time interval for hes_so in 2015 is 0.166666666666667 minutes (0.00277777777777778 hours)"
## [1] "Most common time interval for supsi_ch in 2015 is 43496.6333333333 minutes (724.943888888889 hours)"
## [1] "Most common time interval for FHNW in 2016 is 34708.6666666667 minutes (578.477777777778 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2016 is 240.05 minutes (4.00083333333333 hours)"
## [1] "Most common time interval for ZHAW in 2016 is 21.2 minutes (0.353333333333333 hours)"
## [1] "Most common time interval for bfh in 2016 is 0.0833333333333333 minutes (0.00138888888888889 hours)"
## [1] "Most common time interval for hes_so in 2016 is 2.7 minutes (0.045 hours)"
## [1] "Most common time interval for hslu in 2016 is NA minutes (NA hours)"
## [1] "Most common time interval for supsi_ch in 2016 is 1.58333333333333 minutes (0.0263888888888889 hours)"
## [1] "Most common time interval for FHNW in 2017 is 48748.5333333333 minutes (812.475555555556 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2017 is 5617.83333333333 minutes (93.6305555555555 hours)"
## [1] "Most common time interval for ZHAW in 2017 is 6954.43333333333 minutes (115.907222222222 hours)"
## [1] "Most common time interval for bfh in 2017 is 18606.6666666667 minutes (310.111111111111 hours)"
## [1] "Most common time interval for hes_so in 2017 is 71909.9833333333 minutes (1198.49972222222 hours)"
## [1] "Most common time interval for hslu in 2017 is 0.266666666666667 minutes (0.00444444444444444 hours)"
## [1] "Most common time interval for supsi_ch in 2017 is 1.36666666666667 minutes (0.0227777777777778 hours)"
## [1] "Most common time interval for FHNW in 2018 is 0.166666666666667 minutes (0.00277777777777778 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2018 is 1446.23333333333 minutes (24.1038888888889 hours)"
## [1] "Most common time interval for ZHAW in 2018 is 5689.93333333333 minutes (94.8322222222222 hours)"
## [1] "Most common time interval for bfh in 2018 is 20172.05 minutes (336.200833333333 hours)"
## [1] "Most common time interval for hes_so in 2018 is 31170.8333333333 minutes (519.513888888889 hours)"
## [1] "Most common time interval for hslu in 2018 is 0.233333333333333 minutes (0.00388888888888889 hours)"
## [1] "Most common time interval for supsi_ch in 2018 is 0.183333333333333 minutes (0.00305555555555556 hours)"
## [1] "Most common time interval for FHNW in 2019 is 315.233333333333 minutes (5.25388888888889 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2019 is 10079.85 minutes (167.9975 hours)"
## [1] "Most common time interval for ZHAW in 2019 is 1255.61666666667 minutes (20.9269444444444 hours)"
## [1] "Most common time interval for bfh in 2019 is 1440.05 minutes (24.0008333333333 hours)"
## [1] "Most common time interval for hes_so in 2019 is 1140.03333333333 minutes (19.0005555555556 hours)"
## [1] "Most common time interval for hslu in 2019 is 1.95 minutes (0.0325 hours)"
## [1] "Most common time interval for supsi_ch in 2019 is 15 minutes (0.25 hours)"
## [1] "Most common time interval for FHNW in 2020 is 3180.16666666667 minutes (53.0027777777778 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2020 is 2880.03333333333 minutes (48.0005555555556 hours)"
## [1] "Most common time interval for ZHAW in 2020 is 13693.7666666667 minutes (228.229444444444 hours)"
## [1] "Most common time interval for bfh in 2020 is 14531.7333333333 minutes (242.195555555556 hours)"
## [1] "Most common time interval for hes_so in 2020 is 1139.91666666667 minutes (18.9986111111111 hours)"
## [1] "Most common time interval for hslu in 2020 is 120 minutes (2 hours)"
## [1] "Most common time interval for ost_fh in 2020 is NA minutes (NA hours)"
## [1] "Most common time interval for supsi_ch in 2020 is 0.133333333333333 minutes (0.00222222222222222 hours)"
## [1] "Most common time interval for FHNW in 2021 is 0.5 minutes (0.00833333333333333 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2021 is 0.333333333333333 minutes (0.00555555555555555 hours)"
## [1] "Most common time interval for ZHAW in 2021 is 13043.9833333333 minutes (217.399722222222 hours)"
## [1] "Most common time interval for bfh in 2021 is 1411.05 minutes (23.5175 hours)"
## [1] "Most common time interval for hes_so in 2021 is 0 minutes (0 hours)"
## [1] "Most common time interval for hslu in 2021 is 0 minutes (0 hours)"
## [1] "Most common time interval for ost_fh in 2021 is 0.35 minutes (0.00583333333333333 hours)"
## [1] "Most common time interval for supsi_ch in 2021 is 1140 minutes (19 hours)"
## [1] "Most common time interval for FHNW in 2022 is 1439.93333333333 minutes (23.9988888888889 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2022 is 0.1 minutes (0.00166666666666667 hours)"
## [1] "Most common time interval for ZHAW in 2022 is 18623.7166666667 minutes (310.395277777778 hours)"
## [1] "Most common time interval for bfh in 2022 is 7192.96666666667 minutes (119.882777777778 hours)"
## [1] "Most common time interval for hes_so in 2022 is 5798.53333333333 minutes (96.6422222222222 hours)"
## [1] "Most common time interval for hslu in 2022 is 0 minutes (0 hours)"
## [1] "Most common time interval for ost_fh in 2022 is 0.133333333333333 minutes (0.00222222222222222 hours)"
## [1] "Most common time interval for supsi_ch in 2022 is 28800.7333333333 minutes (480.012222222222 hours)"
## [1] "Most common time interval for FHNW in 2023 is 9997.63333333333 minutes (166.627222222222 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2023 is 21962.3833333333 minutes (366.039722222222 hours)"
## [1] "Most common time interval for ZHAW in 2023 is 70740.3333333333 minutes (1179.00555555556 hours)"
## [1] "Most common time interval for bfh in 2023 is 8000.11666666667 minutes (133.335277777778 hours)"
## [1] "Most common time interval for hes_so in 2023 is 4621.1 minutes (77.0183333333333 hours)"
## [1] "Most common time interval for hslu in 2023 is 627.083333333333 minutes (10.4513888888889 hours)"
## [1] "Most common time interval for supsi_ch in 2023 is 7199 minutes (119.983333333333 hours)"
## [1] "Most common time interval for FH_Graubuenden in 2009 is NA minutes (NA hours)"
## [1] "Most common time interval for FH_Graubuenden in 2010 is 55732.2833333333 minutes (928.871388888889 hours)"
## [1] "Most common time interval for hes_so in 2010 is NA minutes (NA hours)"Overall Trends:
Most Common Time Interval for Tweets: The time intervals between tweets also vary widely among the universities, with some universities having very frequent posts (every few minutes) while others have longer intervals (several hours or even days). Most Active Hour Overall: Across all universities, 11 AM is the most common hour for tweets, with a total of 2356 tweets posted at this time. ### Conclusion: The data indicates that Swiss Universities of Applied Sciences primarily tweet during working hours and show distinct patterns in their most active days and hours. Workers tend to be more productive and active on Twitter in the morning, with a noticeable decline in activity around midday and towards the end of the week. This data-driven approach to analyzing Twitter activity can help universities optimize their social media strategies by identifying the best times and days to engage their audiences.
Question 2: What are the tweets about and how do other Twitter users react to them (likes, etc.)?
Data Preprocessing
langs <- c("de", "fr", "it", "en")
tweets_filtered <- tweets %>%
filter(lang %in% langs)
# Define extended stopwords (outside loop for efficiency)
# Remove 'amp' as it is not meaningful because its only & symbol
# Remove 'rt' because it is an word e.g 'engagiert'.
extended_stopwords <- c(
"#fhnw", "#bfh", "@htw_chur", "#hslu", "#supsi", "#sups",
"amp", "rt", "fr", "ber", "t.co", "https", "http", "www", "com", "html"
)
# Create separate DFMs for each language
dfm_list <- list()
for (sel_lang in langs) {
# Subset tweets for the current language
tweets_lang <- tweets_filtered %>%
filter(lang == sel_lang)
# Create tokens for the current language
stopwords_lang <- stopwords(sel_lang)
# Create tokens for all tweets:
# - create corpus and tokens because tokensonly works on character, corpus, list, tokens, tokens_xptr objects.
# - create tokens and remove: URLS, Punctuation, Numbers, Symbols, Separators
# - transform to lowercase
# - Stem all words
# - Create n-grams of any length (not includinf bigrams and trigrams but they are shown later)
# - It is important to remove the stopwords after stemming the words because we remove the endings from some stem words
tokens_lang <- tweets_lang %>%
corpus(text_field = "full_text_emojis") %>%
tokens(
remove_punct = TRUE, remove_symbols = TRUE, remove_numbers = TRUE,
remove_url = TRUE, remove_separators = TRUE
) %>%
tokens_tolower() %>%
tokens_wordstem(lang = sel_lang) %>%
tokens_ngrams(n = 1) %>%
tokens_select(
pattern =
c(stopwords_lang, extended_stopwords), selection = "remove"
)
# Create DFM for the current language
dfm_list[[sel_lang]] <- dfm(tokens_lang)
}Content Analysis
Tweets were analyzed across four languages: German, French, Italian, and English. Each university tends to tweet predominantly in one or more languages, reflecting the linguistic diversity of Switzerland.
- German: Predominantly used by BFH and FHNW. Common words include “neu” (new), “mehr” (more), “schweiz” (Switzerland), and “studier” (study).
- French: Primarily used by HES-SO. Common words include “projet” (project), “recherch” (research), and “tudi” (study).
- Italian: Mostly used by SUPSI. Common words include “nuov” (new), “progett” (project), and “student” (student).
- English: Frequently used by HSLU. Common words include “student”, “project”, “thank”, and “university”. ### Word Frequency:
- English: The most frequent words were “student”, “new”, “@hslu”, “university”, “project”, “thank”, “@zhaw”, “day”, “science”, and “today”.
- German: The most frequent words were “neu”, “mehr”, “schweiz”, “werd”, “all”, “studier”, “heut”, “hochschul”, “bfh”, and “jahr”.
- Italian: The most frequent words were “nuov”, “sups”, “progett”, “student”, “present”, “info”, “iscrizion”, “cors”, “ricerc”, and “formazion”.
- French: The most frequent words were “hes-so”, “right”, “arrow”, “dan”, “projet”, “a”, “tudi”, “haut”, “col”, and “@hes_so”.
It’s important to note that some words like “right” 👉 and “arrow” ➡️ are actually names of parsed emojis and not written words in the tweets.
Word clouds for each language visually depicted the most common words, emphasizing their relative frequencies. The analysis revealed that universities tweet in multiple languages, reflecting the linguistic diversity of their audience. The most common words often related to educational themes, projects, and institutional news, indicating a focus on academic content.
# Word Frequencies & Visualization
words_freqs_en <- sort(colSums(dfm_list$en), decreasing = TRUE)
head(words_freqs_en, 20)## student new @hslu univers project thank
## 106 74 70 62 60 60
## @zhaw day scienc today innov now
## 59 56 54 52 51 50
## swiss switzerland @fhnw great us join
## 49 49 46 46 44 43
## studi research
## 42 42wordcloud2(data.frame(
word = names(words_freqs_en),
freq = words_freqs_en
), size = 0.5)words_freqs_de <- sort(colSums(dfm_list$de), decreasing = TRUE)
head(words_freqs_de, 20)## neu mehr schweiz werd all studier heut hochschul
## 1586 1104 967 772 706 706 638 601
## bfh jahr knnen digital thema studi projekt welch
## 577 535 507 499 497 466 465 462
## bern statt zeigt arbeit
## 454 451 437 434wordcloud2(data.frame(
word = names(words_freqs_de),
freq = words_freqs_de
), size = 0.5)word_freqs_it <- sort(colSums(dfm_list$it), decreasing = TRUE)
head(word_freqs_it, 20)## nuov sups progett student present info
## 210 208 173 146 143 143
## iscrizion cors ricerc formazion #supsinews #supsievent
## 142 141 135 134 134 129
## scopr inform diplom bachelor apert tutt
## 123 120 116 111 110 105
## master pi
## 103 102wordcloud2(data.frame(
word = names(word_freqs_it),
freq = word_freqs_it
), size = 0.5)# It seems that there are some english words but I think this are emojis
word_freqs_fr <- sort(colSums(dfm_list$fr), decreasing = TRUE)
head(word_freqs_fr, 20)## hes-so right arrow dan projet a tudi haut
## 505 432 324 249 248 234 199 183
## col @hes_so @hessoval dcouvr book open recherch #hes_so
## 155 140 129 127 123 118 117 115
## suiss plus mast nouveau
## 110 105 103 98wordcloud2(data.frame(
word = names(word_freqs_fr),
freq = word_freqs_fr
), size = 0.5)# University-specific Analysis
for (uni in unique(tweets$university)) {
# Subset tweets for the current language
uni_tweets <- tweets_filtered %>%
filter(university == uni)
tokens_lang <- uni_tweets %>%
corpus(text_field = "full_text_emojis") %>%
tokens(
remove_punct = TRUE, remove_symbols = TRUE, remove_numbers = TRUE,
remove_url = TRUE, remove_separators = TRUE
) %>%
tokens_tolower() %>%
tokens_wordstem() %>%
tokens_ngrams(n = 1) %>%
tokens_select(
pattern =
c(
stopwords("en"), stopwords("de"),
stopwords("fr"), stopwords("it"), extended_stopwords
), selection = "remove"
)
# Create Data Frame Matrix for uni with all languages
uni_dfm <- dfm(tokens_lang)
# Word Frequencies
uni_word_freqs <- sort(colSums(uni_dfm), decreasing = TRUE)
# print most common words: the emoji right are used often
head(uni_word_freqs, 20)
wordcloud2(data.frame(
word = names(uni_word_freqs),
freq = uni_word_freqs
), size = 0.5)
}Userreaction Analysis
A weighted engagement metric was calculated to measure user reactions, considering both likes (favorites) and retweets, with retweets given double weight.
Posting Times of Most Engaged Tweets: The analysis of the posting times of the most engaged tweets (top 1000 by engagement) showed that:
- The most engaged tweets were posted throughout the day, with a notable peak around mid-morning (11 AM).
- This pattern aligns with the overall finding that users tend to be more active and engaged during typical working hours. Which can be a sign, that the users are more active during the day and not during the night. Or the customers of the universities are working also in a university or are students.
Content of Most Engaged Tweets:
The most common words in the most engaged tweets included “mehr” (more), “neue” (new), “schweiz” (Switzerland), “schweizer” (Swiss), “right”, “heut” (today), “zeigt” (shows), “#hsluinformatik” (HSLU informatics), “studi” (study), and “zhaw”. Again, “right” and similar terms are names of emojis and not actual
# Calculate a 'weighted engagement' metric
tweets <- tweets %>%
mutate(
weighted_engagement = favorite_count * 1 + retweet_count * 2
)
# Identify tweets with the highest weighted engagement
most_engaged_tweets <- tweets %>%
arrange(desc(weighted_engagement)) %>%
head(1000) # Top 1000 for analysis
# Analyze posting time of most engaged tweets (same as before)
most_engaged_tweets_time <- most_engaged_tweets %>%
mutate(time_of_day = format(created_at, "%H"))
ggplot(most_engaged_tweets_time, aes(x = as.numeric(time_of_day))) +
geom_histogram(binwidth = 1, fill = "lightblue", color = "blue") +
labs(
title = "Distribution of Posting Times for Most Engaged Tweets",
x = "Hour of Day",
y = "Frequency"
)
Analyse the content of the most liked tweets
# Preprocessing content of most liked tweets
tokens_most_engaged <- most_engaged_tweets %>%
corpus(text_field = "full_text_emojis") %>%
tokens(
remove_punct = TRUE, remove_symbols = TRUE, remove_numbers = TRUE,
remove_url = TRUE, remove_separators = TRUE
) %>%
tokens_tolower() %>%
tokens_wordstem(lang = sel_lang) %>%
tokens_ngrams(n = 1) %>%
tokens_select(
pattern =
c(
stopwords("en"), stopwords("de"),
stopwords("fr"), stopwords("it"), extended_stopwords
), selection = "remove"
)
tokens_most_engaged_dfm <- dfm(tokens_most_engaged)
freqs_most_engaged <- sort(colSums(tokens_most_engaged_dfm), decreasing = TRUE)
# print most common words: the emoji right are used often
head(freqs_most_engaged, 20)## mehr neue schweiz schweizer right
## 81 67 48 47 46
## heut zeigt #hsluinformatik studi zhaw
## 44 41 40 39 39
## hes-so knnen neuen hochschul campus
## 38 38 36 34 33
## innov gibt ab entwickelt bfh
## 31 30 30 30 30set.seed(123)
wordcloud2(data.frame(
word = names(freqs_most_engaged),
freq = freqs_most_engaged
), size = 0.5)The analysis indicates that Swiss Universities of Applied Sciences tweet in multiple languages, reflecting the linguistic diversity of their audience. The tweets often focus on educational themes, projects, and institutional news. User engagement is highest for tweets posted during working hours, with the most engaging content often including timely updates and relevant academic information. Recognizing the role of emojis in enhancing engagement, universities can further optimize their social media strategies to maximize reach and impact. ## Question 3: How do the university tweets differ in terms of content, style, emotions, etc?
Content Analysis (Word Clouds)
Each university shows distinct patterns in the words and emojis used in their tweets. The analysis involved creating word clouds and identifying the most common words and emojis.
Most Common Words:
- FHNW: Common words include “mehr” (more), “hochschul” (university), and “studierend” (students).
- FH Graubünden: Words like “chur” (location), “htw” (university abbreviation), and “busi” (business) are frequent.
- ZHAW: Frequent words include “zhaw” (university abbreviation), “engineering”, and “schweizer” (Swiss).
- BFH: Common terms are “bern” (location), “projekt” (project), and “hochschul” (university).
- HES-SO: Words like “hes-so” (university abbreviation), “projet” (project), and “tudiant” (student) are prevalent.
- HSLU: Common words are “hslu” (university abbreviation), “luzern” (location), and “hochschul” (university).
- OST-FH: Frequent terms include “ost” (university abbreviation), “st.gallen” (location), and “podcast”.
- SUPSI-CH: Words like “supsi” (university abbreviation), “formazion” (education), and “progetto” (project) are prevalent.
Most Common Emojis: - FHNW: Top emojis include 👉 (backhand index pointing right), 💛 (yellow heart), and 🖤 (black heart). - FH Graubünden: Frequent emojis are 🎉 (party popper), 😃 (grinning face with big eyes), and 😊 (blush). - ZHAW: Common emojis include 👉 (backhand index pointing right), ⚡ (high voltage), and 😉 (wink). - BFH: Top emojis are 👉 (backhand index pointing right), 🔋 (battery), and 👇 (backhand index pointing down). - HES-SO: Common emojis are 👉 (backhand index pointing right), 🎓 (graduation cap), and ➡ (arrow right). - HSLU: Top emojis include 🎓 (graduation cap), 👨 (man), and 🚀 (rocket). - OST-FH: Frequent emojis are 👉 (backhand index pointing right), ➡ (arrow right), and 🎓 (graduation cap). - SUPSI-CH: Common emojis include 👉 (backhand index pointing right), 🎓 (graduation cap), and 🎉 (party popper).
for (uni in unique(tweets$university)) {
uni_tweets <- tweets %>%
filter(university == uni, lang %in% langs)
tokens_uni <- uni_tweets %>%
corpus(text_field = "full_text_emojis") %>%
tokens(
remove_punct = TRUE, remove_symbols = TRUE, remove_numbers = TRUE,
remove_url = TRUE, remove_separators = TRUE
) %>%
tokens_tolower() %>%
tokens_wordstem() %>%
tokens_ngrams(n = 1) %>%
tokens_select(
pattern =
c(
stopwords("en"), stopwords("de"),
stopwords("fr"), stopwords("it"), extended_stopwords
), selection = "remove"
)
uni_dfm <- dfm(tokens_uni)
freqs_uni <- sort(colSums(uni_dfm), decreasing = TRUE)
# print most common words: the emoji right are used often
head(freqs_uni, 20)
set.seed(123)
wordcloud2(data.frame(
word = names(freqs_uni),
freq = freqs_uni
), size = 0.5)
# Analyze Top Emojis by University
emoji_count_per_university <- uni_tweets %>%
top_n_emojis(full_text)
print(emoji_count_per_university)
emoji_count_per_university %>%
mutate(emoji_name = reorder(emoji_name, n)) %>%
ggplot(aes(n, emoji_name)) +
geom_col() +
labs(x = "Count", y = NULL, title = "Top 20 Emojis Used")
}## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 backhand_index_pointing_right 👉 People & Body 56
## 2 yellow_heart 💛 Smileys & Emotion 34
## 3 black_heart 🖤 Smileys & Emotion 32
## 4 woman 👩 People & Body 28
## 5 man 👨 People & Body 17
## 6 clap 👏 People & Body 16
## 7 flag_Switzerland 🇨🇭 Flags 15
## 8 microscope 🔬 Objects 15
## 9 computer 💻 Objects 14
## 10 graduation_cap 🎓 Objects 13
## 11 school 🏫 Travel & Places 13
## 12 face_with_medical_mask 😷 Smileys & Emotion 12
## 13 raised_hands 🙌 People & Body 12
## 14 robot 🤖 Smileys & Emotion 12
## 15 female_sign ♀️ Symbols 10
## 16 trophy 🏆 Activities 9
## 17 woman_scientist 👩🔬 People & Body 9
## 18 party_popper 🎉 Activities 8
## 19 star_struck 🤩 Smileys & Emotion 8
## 20 sun_with_face 🌞 Travel & Places 8
## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 party_popper 🎉 Activities 18
## 2 grinning_face_with_big_eyes 😃 Smileys & Emotion 15
## 3 blush 😊 Smileys & Emotion 8
## 4 smiling_face_with_sunglasses 😎 Smileys & Emotion 8
## 5 bulb 💡 Objects 7
## 6 +1 👍 People & Body 6
## 7 camera_flash 📸 Objects 6
## 8 flexed_biceps 💪 People & Body 6
## 9 four_leaf_clover 🍀 Animals & Nature 6
## 10 grinning_face_with_smiling_eyes 😄 Smileys & Emotion 6
## 11 heart_eyes 😍 Smileys & Emotion 6
## 12 hugs 🤗 Smileys & Emotion 6
## 13 female_sign ♀️ Symbols 4
## 14 graduation_cap 🎓 Objects 4
## 15 grinning 😀 Smileys & Emotion 4
## 16 robot 🤖 Smileys & Emotion 4
## 17 backhand_index_pointing_down 👇 People & Body 3
## 18 computer 💻 Objects 3
## 19 lady_beetle 🐞 Animals & Nature 3
## 20 ocean 🌊 Travel & Places 3
## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 backhand_index_pointing_right 👉 People & Body 21
## 2 high_voltage ⚡ Travel & Places 11
## 3 wink 😉 Smileys & Emotion 9
## 4 clap 👏 People & Body 5
## 5 flag_Switzerland 🇨🇭 Flags 5
## 6 rocket 🚀 Travel & Places 5
## 7 +1 👍 People & Body 4
## 8 arrow_right ➡️ Symbols 4
## 9 bug 🐛 Animals & Nature 3
## 10 computer 💻 Objects 3
## 11 flexed_biceps 💪 People & Body 3
## 12 man 👨 People & Body 3
## 13 bangbang ‼️ Symbols 2
## 14 dark_skin_tone 🏿 Component 2
## 15 exclamation ❗ Symbols 2
## 16 female_sign ♀️ Symbols 2
## 17 four_leaf_clover 🍀 Animals & Nature 2
## 18 green_salad 🥗 Food & Drink 2
## 19 grinning 😀 Smileys & Emotion 2
## 20 medium_light_skin_tone 🏼 Component 2
## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 backhand_index_pointing_right 👉 People & Body 49
## 2 battery 🔋 Objects 16
## 3 backhand_index_pointing_down 👇 People & Body 12
## 4 woman 👩 People & Body 12
## 5 palm_tree 🌴 Animals & Nature 11
## 6 bulb 💡 Objects 10
## 7 computer 💻 Objects 10
## 8 evergreen_tree 🌲 Animals & Nature 10
## 9 graduation_cap 🎓 Objects 10
## 10 party_popper 🎉 Activities 10
## 11 robot 🤖 Smileys & Emotion 10
## 12 clap 👏 People & Body 9
## 13 coconut 🥥 Food & Drink 9
## 14 date 📅 Objects 9
## 15 deciduous_tree 🌳 Animals & Nature 9
## 16 flag_Switzerland 🇨🇭 Flags 9
## 17 rocket 🚀 Travel & Places 9
## 18 automobile 🚗 Travel & Places 8
## 19 clinking_glasses 🥂 Food & Drink 8
## 20 seedling 🌱 Animals & Nature 8
## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 arrow_right ➡️ Symbols 320
## 2 arrow_heading_down ⤵️ Symbols 245
## 3 book 📖 Objects 115
## 4 mag_right 🔎 Objects 97
## 5 mega 📣 Objects 53
## 6 clapper 🎬 Objects 38
## 7 NEW_button 🆕 Symbols 35
## 8 computer 💻 Objects 35
## 9 microscope 🔬 Objects 32
## 10 bulb 💡 Objects 29
## 11 police_car_light 🚨 Travel & Places 27
## 12 backhand_index_pointing_right 👉 People & Body 26
## 13 graduation_cap 🎓 Objects 23
## 14 studio_microphone 🎙️ Objects 23
## 15 clap 👏 People & Body 21
## 16 date 📅 Objects 17
## 17 medal_sports 🏅 Activities 15
## 18 memo 📝 Objects 15
## 19 woman 👩 People & Body 15
## 20 flag_Switzerland 🇨🇭 Flags 14
## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 sparkles ✨ Activities 28
## 2 flag_Switzerland 🇨🇭 Flags 18
## 3 rocket 🚀 Travel & Places 12
## 4 party_popper 🎉 Activities 11
## 5 partying_face 🥳 Smileys & Emotion 9
## 6 Christmas_tree 🎄 Activities 7
## 7 clap 👏 People & Body 7
## 8 star ⭐ Travel & Places 7
## 9 bottle_with_popping_cork 🍾 Food & Drink 6
## 10 bulb 💡 Objects 5
## 11 glowing_star 🌟 Travel & Places 5
## 12 smiling_face_with_sunglasses 😎 Smileys & Emotion 5
## 13 +1 👍 People & Body 4
## 14 camera_flash 📸 Objects 4
## 15 clinking_glasses 🥂 Food & Drink 4
## 16 four_leaf_clover 🍀 Animals & Nature 4
## 17 musical_notes 🎶 Objects 4
## 18 person_running 🏃 People & Body 4
## 19 raised_hands 🙌 People & Body 4
## 20 robot 🤖 Smileys & Emotion 4
## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 graduation_cap 🎓 Objects 3
## 2 man 👨 People & Body 2
## 3 man_student 👨🎓 People & Body 2
## 4 rocket 🚀 Travel & Places 2
## 5 snowflake ❄️ Travel & Places 2
## 6 backhand_index_pointing_right 👉 People & Body 1
## 7 brain 🧠 People & Body 1
## 8 chocolate_bar 🍫 Food & Drink 1
## 9 clapper 🎬 Objects 1
## 10 eyes 👀 People & Body 1
## 11 fire 🔥 Travel & Places 1
## 12 flexed_biceps 💪 People & Body 1
## 13 grinning 😀 Smileys & Emotion 1
## 14 heart_eyes_cat 😻 Smileys & Emotion 1
## 15 high_voltage ⚡ Travel & Places 1
## 16 mantelpiece_clock 🕰️ Travel & Places 1
## 17 sleeping 😴 Smileys & Emotion 1
## 18 slightly_smiling_face 🙂 Smileys & Emotion 1
## 19 sun ☀️ Travel & Places 1
## 20 woman 👩 People & Body 1
## # A tibble: 20 × 4
## emoji_name unicode emoji_category n
## <chr> <chr> <chr> <int>
## 1 arrow_right ➡️ Symbols 83
## 2 backhand_index_pointing_right 👉 People & Body 21
## 3 graduation_cap 🎓 Objects 19
## 4 arrow_forward ▶️ Symbols 18
## 5 bulb 💡 Objects 10
## 6 rocket 🚀 Travel & Places 9
## 7 party_popper 🎉 Activities 8
## 8 flag_Switzerland 🇨🇭 Flags 7
## 9 clap 👏 People & Body 6
## 10 exclamation ❗ Symbols 5
## 11 SOON_arrow 🔜 Symbols 4
## 12 grinning_face_with_big_eyes 😃 Smileys & Emotion 4
## 13 camera_flash 📸 Objects 3
## 14 computer 💻 Objects 3
## 15 movie_camera 🎥 Objects 3
## 16 rainbow 🌈 Travel & Places 3
## 17 studio_microphone 🎙️ Objects 3
## 18 woman 👩 People & Body 3
## 19 Christmas_tree 🎄 Activities 2
## 20 backhand_index_pointing_down 👇 People & Body 2# Generate general tokens for bigram and trigram analysis
tokens <- tweets %>%
corpus(text_field = "full_text_emojis") %>%
tokens(
remove_punct = TRUE, remove_symbols = TRUE, remove_numbers = TRUE,
remove_url = TRUE, remove_separators = TRUE
) %>%
tokens_tolower() %>%
tokens_wordstem() %>%
tokens_select(
pattern =
c(
stopwords("en"), stopwords("de"),
stopwords("fr"), stopwords("it"), extended_stopwords
), selection = "remove"
)
# Bigram Wordcloud
bi_gram_tokens <- tokens_ngrams(tokens, n = 2)
dfm_bi_gram <- dfm(bi_gram_tokens)
freqs_bi_gram <- sort(colSums(dfm_bi_gram), decreasing = TRUE)
head(freqs_bi_gram, 20)## right_arrow htw_chur index_point
## 421 259 207
## backhand_index hochschul_luzern point_right
## 206 185 183
## berner_fachhochschul sozial_arbeit prof_dr
## 157 154 142
## haut_cole herzlich_gratul open_book
## 141 139 117
## magnifi_glass glass_tilt tilt_right
## 97 97 97
## fh_graubnden neusten_blogbeitrag book_#revuehmisphr
## 91 87 85
## social_media advanc_studi
## 84 83# Create the bigram word cloud
set.seed(123)
wordcloud2(data.frame(
word = names(freqs_bi_gram),
freq = freqs_bi_gram
), size = 0.5)# Trigram Wordcloud
tri_gram_tokens <- tokens_ngrams(tokens, n = 3)
dfm_tri_gram <- dfm(tri_gram_tokens)
reqs_tri_gram <- sort(colSums(dfm_tri_gram), decreasing = TRUE)
head(reqs_tri_gram, 20)## backhand_index_point index_point_right
## 206 183
## magnifi_glass_tilt glass_tilt_right
## 97 97
## open_book_#revuehmisphr hochschul_gestaltung_kunst
## 85 62
## dipartimento_tecnologi_innov master_advanc_studi
## 40 38
## depart_sozial_arbeit #infoanlass_mrz_findet
## 36 33
## polic_car_light univers_appli_scienc
## 32 31
## busi_administr_statt findet_#zrich_infoanlass
## 30 30
## tag_offenen_tr hochschul_life_scienc
## 29 29
## gestaltung_kunst_fhnw mas_busi_administr
## 29 28
## mehr_neuen_blogbeitrag mehr_neusten_blogbeitrag
## 28 28# Create the bigram word cloud
set.seed(123)
wordcloud2(data.frame(
word = names(reqs_tri_gram),
freq = reqs_tri_gram
), size = 0.5)LDA Topic Modeling
# Source: Christoph Zangger -> löscht alle Reihen mit nur 0s
new_dfm <- dfm_subset(dfm_list$en, ntoken(dfm_list$en) > 0)
tweet_lda <- LDA(new_dfm, k = 5, control = list(seed = 123))
# Tidy the LDA results
topic_terms <- tidy(tweet_lda, matrix = "beta")
# Extract topics and top terms
topics <- as.data.frame(terms(tweet_lda, 50)) # First fifty words per topic
# Extract top terms per topic
top_terms <- topic_terms %>%
group_by(topic) %>%
top_n(8, beta) %>% # Show top 8 terms per topic
ungroup() %>%
arrange(topic, -beta)
# Visualize top terms per topic
top_terms %>%
mutate(term = reorder_within(term, beta, topic)) %>%
ggplot(aes(beta, term, fill = factor(topic))) +
geom_col(show.legend = FALSE) +
facet_wrap(~topic, scales = "free") +
scale_y_reordered() +
labs(
x = "Beta (Term Importance within Topic)",
y = NULL,
title = "Top Terms per Topic in Tweets (LDA)"
)
# Most different words among topics (using log ratios)
diff <- topic_terms %>%
mutate(topic = paste0("topic", topic)) %>%
spread(topic, beta) %>%
filter(topic1 > .001 | topic2 > .001 | topic3 > .001) %>%
mutate(
logratio_t1t2 = log2(topic2 / topic1),
logratio_t1t3 = log2(topic3 / topic1),
logratio_t2t3 = log2(topic3 / topic2)
)
diff## # A tibble: 328 × 9
## term topic1 topic2 topic3 topic4 topic5 logratio_t1t2 logratio_t1t3
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 @academi… 1.96e-3 5.43e-4 1.53e-3 3.59e-3 3.22e-3 -1.86 -0.358
## 2 @bfh_hesb 1.60e-3 3.51e-3 2.70e-3 4.10e-3 3.54e-3 1.13 0.757
## 3 @ch_univ… 3.20e-4 1.28e-3 6.28e-4 1.06e-4 1.67e-4 2.01 0.975
## 4 @fh_grau… 1.47e-3 4.17e-4 1.24e-4 8.18e-4 1.76e-3 -1.82 -3.58
## 5 @fhnw 2.59e-3 7.56e-3 7.94e-4 1.15e-3 7.12e-3 1.55 -1.70
## 6 @fhnwbusi 5.02e-3 3.18e-3 2.27e-3 5.57e-3 6.68e-4 -0.659 -1.15
## 7 @globalc… 1.89e-3 4.31e-4 2.40e-4 2.90e-4 7.17e-5 -2.13 -2.97
## 8 @greater… 1.96e-3 1.90e-4 2.21e-3 8.21e-4 2.32e-4 -3.37 0.172
## 9 @grstift… 9.60e-4 2.17e-3 1.41e-3 1.92e-3 2.31e-3 1.17 0.559
## 10 @hes_so 4.52e-4 1.27e-3 3.09e-3 9.90e-4 2.13e-3 1.49 2.77
## # ℹ 318 more rows
## # ℹ 1 more variable: logratio_t2t3 <dbl># LDA Topic Modeling for each university
universities <- unique(tweets$university)
for (uni in universities) {
# Filter tweets for the current university
uni_tweets <- tweets %>% filter(university == uni)
tokens_uni <- uni_tweets %>%
corpus(text_field = "full_text_emojis") %>%
tokens(
remove_punct = TRUE, remove_symbols = TRUE, remove_numbers = TRUE,
remove_url = TRUE, remove_separators = TRUE
) %>%
tokens_tolower() %>%
tokens_wordstem() %>%
tokens_ngrams(n = 1) %>%
tokens_select(
pattern =
c(
stopwords("en"), stopwords("de"),
stopwords("fr"), stopwords("it"), extended_stopwords
), selection = "remove"
)
uni_dfm <- dfm(tokens_uni)
# Apply LDA
uni_dfm <- dfm_subset(uni_dfm, ntoken(uni_dfm) > 0)
tweet_lda <- LDA(uni_dfm, k = 5, control = list(seed = 123))
# Tidy the LDA results
tweet_lda_td <- tidy(tweet_lda)
# Extract top terms per topic
top_terms <- tweet_lda_td %>%
group_by(topic) %>%
top_n(8, beta) %>%
ungroup() %>%
arrange(topic, -beta)
# Visualize top terms per topic
p <- top_terms %>%
mutate(term = reorder_within(term, beta, topic)) %>%
ggplot(aes(beta, term, fill = factor(topic))) +
geom_col(show.legend = FALSE) +
facet_wrap(~topic, scales = "free") +
scale_y_reordered() +
labs(
x = "Beta (Term Importance within Topic)",
y = NULL,
title = paste("Top Terms per Topic in Tweets from", uni, "(LDA)")
)
print(p)
# Topic Model Summary: top 10 terms per topic
cat("\nTopic Model Summary for", uni, ":\n")
print(as.data.frame(terms(tweet_lda, 10)))
}
##
## Topic Model Summary for FHNW :
## Topic 1 Topic 2 Topic 3 Topic 4 Topic 5
## 1 @hsafhnw @fhnwbusi fhnw fhnw @fhnwbusi
## 2 fhnw @hsafhnw @fhnw @fhnwbusi @fhnwtechnik
## 3 swiss mehr campus hochschul @fhnw
## 4 challeng fhnw mehr @fhnwtechnik @hsafhnw
## 5 morgen hochschul heut @fhnwpsychologi mehr
## 6 neue @fhnwtechnik studierend projekt schweiz
## 7 brugg-windisch studierend hochschul neue neuen
## 8 mehr prof ab index kunst
## 9 olten dr @hsafhnw basel campus
## 10 erklrt brugg-windisch neue @fhnw heut
##
## Topic Model Summary for FH_Graubuenden :
## Topic 1 Topic 2 Topic 3 Topic 4 Topic 5
## 1 chur blogbeitrag htw statt statt
## 2 #infoanlass mehr #htwchur #htwchur chur
## 3 #htwchur neuen chur findet findet
## 4 htw infoanlass busi onlin #htwchur
## 5 blogbeitrag findet neusten htw #fhgr
## 6 graubnden #fhgr #studium manag mehr
## 7 #fhgr graubnden graubnden heut graubnden
## 8 #graubnden chur manag mehr @suedostschweiz
## 9 thema @htwchurtour infoanlass product htw
## 10 #chur fh studium #chur blogbeitrag
##
## Topic Model Summary for ZHAW :
## Topic 1 Topic 2 Topic 3 Topic 4
## 1 zhaw @zhaw #zhaw @iam_winterthur
## 2 dank @engineeringzhaw winterthur schweiz
## 3 @engineeringzhaw neue heut knnen
## 4 @zhaw @sml_zhaw neue mehr
## 5 winterthur studierend @engineeringzhaw #zhaw
## 6 thema schweizer cc neue
## 7 schweizer #zhawimpact gibt cc
## 8 zeigt @c_caviglia #zhawimpact neuen
## 9 heut via via studi
## 10 via cc @iam_winterthur schweizer
## Topic 5
## 1 zhaw
## 2 @engineeringzhaw
## 3 cc
## 4 mehr
## 5 zeigt
## 6 heut
## 7 @sml_zhaw
## 8 knnen
## 9 studi
## 10 gibt
##
## Topic Model Summary for bfh :
## Topic 1 Topic 2 Topic 3 Topic 4 Topic 5
## 1 bfh bfh bern bfh mehr
## 2 biel thema neue neue bfh
## 3 bern berner berner @bfh_hesb #knoten_maschen
## 4 mehr @bfh_hesb mehr bern bern
## 5 arbeit fachhochschul innen thema berner
## 6 projekt @hkb_bfh neuen zukunft schweizer
## 7 @hkb_bfh ab zeigt projekt @bfh_hesb
## 8 fachhochschul erfahren thema digital knnen
## 9 zeigt biel studi geht anmelden
## 10 index statt nachhaltig neu neue
##
## Topic Model Summary for hes_so :
## Topic 1 Topic 2 Topic 3 Topic 4 Topic 5
## 1 arrow hes-so hes-so @hes_so right
## 2 dan right haut projet projet
## 3 right arrow cole master arrow
## 4 projet @hes_so tudiant tudiant @hessovalai
## 5 book dan dan open haut
## 6 suiss tilt arrow @hessovalai professeur
## 7 #revuehmisphr #hes_so #hes_so right domain
## 8 nouvell recherch plus hes-so master
## 9 @hessovalai travail programm diplm open
## 10 dcouvrez glass recherch recherch nouveau
##
## Topic Model Summary for hslu :
## Topic 1 Topic 2 Topic 3 Topic 4
## 1 @hslu hochschul #hsluinformatik luzern
## 2 mehr luzern studi mehr
## 3 luzern @hslu interview @hslu
## 4 neue mehr zeigt neue
## 5 #hsluinformatik schweiz depart finden
## 6 depart heut menschen #hslumusik
## 7 design #hsluinformatik #hsluwirtschaft heut
## 8 bachelor projekt design studium
## 9 jahr gibt #hslusozialearbeit #hsluwirtschaft
## 10 schweizer schweizer neuen digitalen
## Topic 5
## 1 @hslu
## 2 zeigt
## 3 welch
## 4 luzern
## 5 heut
## 6 knnen
## 7 depart
## 8 schweizer
## 9 studierend
## 10 digit
##
## Topic Model Summary for ost_fh :
## Topic 1 Topic 2
## 1 #ostschweizerfachhochschul @ozg_ost
## 2 @ost_fh #ostschweizerfachhochschul
## 3 #informatik @ost_fh
## 4 ost ostschweiz
## 5 st.gallen ost
## 6 neu neue
## 7 @eastdigit fachhochschul
## 8 #wirtschaftsinformatik drei
## 9 @itrockt #countdown
## 10 podcast thema
## Topic 3 Topic 4
## 1 ost ost
## 2 #ostschweizerfachhochschul #ostschweizerfachhochschul
## 3 @ost_fh mehr
## 4 rapperswil leben
## 5 bachelor rapperswil
## 6 campus kulturzyklus
## 7 fhs podcast
## 8 st.gallen onlin
## 9 alumni kontrast
## 10 statt @ost_fh
## Topic 5
## 1 @ost_fh
## 2 #ostschweizerfachhochschul
## 3 mehr
## 4 @ost_wi
## 5 ost
## 6 projekt
## 7 schweizer
## 8 neuen
## 9 thema
## 10 @ozg_ost
##
## Topic Model Summary for supsi_ch :
## Topic 1 Topic 2 Topic 3 Topic 4 Topic 5
## 1 arrow supsi supsi #supsiev info
## 2 right #supsinew formazion supsi pi
## 3 supsi studenti progetto bachelor #supsiev
## 4 #supsiev oggi studi formazion manag
## 5 progetto master #supsinew master iscrizioni
## 6 #supsinew deg nuovo tecnologi formazion
## 7 studenti iscrizioni @usi_univers corsi nuovo
## 8 @supsi_ch informazioni @supsi_ch scopri dipartimento
## 9 stream busi campus @usi_univers novembr
## 10 svizzera ingegneria pi @supsi_ch temaStyle Analysis
The distribution of tweet lengths shows variation across universities. Most tweets are concise, aligning with Twitter’s character limit, but the exact length distribution differs among institutions. It is interesting to see that much tweets have around 150 words and that the tweets from the universities are not that long. It is a typical sign that the tweets are not that long and this is a common thing in social media.
tweets %>%
mutate(tweet_length = nchar(full_text)) %>%
ggplot(aes(x = tweet_length)) +
geom_histogram() +
labs(title = "Distribution of Tweet Lengths")
### Sentiment Analysis Sentiment analysis was conducted to evaluate the
emotional tone of the tweets. The analysis used the Syuzhet method to
calculate sentiment scores for each tweet.
Overall Sentiment Trends: - The sentiment scores vary over time and by university, showing fluctuations in the emotional tone of the tweets. - Positive words commonly found in tweets include terms related to academic achievements, collaborations, and positive experiences. - Negative words often relate to challenges, competitions, and issues faced by the universities.
Sentiment by University: - FHNW: Positive words include “academy”, “accelerate”, and “activities”. Negative words include “avoid”, “bacteria”, and “challenge”. - FH Graubünden: Positive words include “able”, “academic”, and “advantage”. Negative words include “competition”, “corruption”, and “fire”. - ZHAW: Positive words include “abilities”, “academic”, and “achievement”. Negative words include “barrier”, “challenge”, and “competition”. - BFH: Positive words include “academic”, “access”, and “activities”. Negative words include “aggression”, “competition”, and “fail”. - HES-SO: Positive words include “academic”, “active”, and “amazing”. Negative words include “confessions”, “failure”, and “hard”. - HSLU: Positive words include “academic”, “access”, and “achievement”. Negative words include “addiction”, “challenge”, and “fail”. - OST-FH: Positive words include “announce”, “beautiful”, and “collaboration”. Negative words are minimal, including “dire” and “fire”. - SUPSI-CH: Positive words include “academic”, “access”, and “achievement”. Negative words include “barrier”, “cloud”, and “danger”.
# Calculate Sentiment for Supported Languages Only
langs <- c("de", "fr", "it", "en")
tweets_filtered <- tweets %>%
filter(lang %in% langs)
# TODO: because sentiment only works for english
# Create Function to Get Syuzhet Sentiment
get_syuzhet_sentiment <- function(text, lang) {
# Check if language is supported
if (lang %in% langs) {
return(get_sentiment(text, method = "syuzhet", lang = lang))
} else {
return(NA) # Return NA for unsupported languages
}
}
# Calculate Syuzhet Sentiment for each Tweet
tweets_filtered$sentiment <-
mapply(get_syuzhet_sentiment, tweets_filtered$full_text, tweets_filtered$lang)
plot_data <- tweets_filtered %>%
group_by(university, month) %>%
summarize(mean_sentiment_syuzhet = mean(sentiment, na.rm = TRUE))
# Plot Syuzhet Sentiment by all Universities
ggplot(plot_data, aes(
x = month,
y = mean_sentiment_syuzhet,
color = university, group = university
)) +
geom_line() +
labs(
title = "Mean Syuzhet Sentiment Over Time by University",
y = "Mean Sentiment Score"
) +
scale_x_datetime(date_breaks = "1 month", date_labels = "%Y-%m") +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
for (uni in unique(tweets$university)) {
uni_tweets <- tweets %>%
filter(university == uni, lang == "en")
uni_tweets$sentiment <-
mapply(get_syuzhet_sentiment, uni_tweets$full_text, uni_tweets$lang)
plot_data <- uni_tweets %>%
group_by(month) %>%
summarize(mean_sentiment = mean(sentiment, na.rm = TRUE))
# Plot Syuzhet Sentiment Over Time (Per University)
print(ggplot(plot_data, aes(x = month, y = mean_sentiment, group = 1)) +
geom_line() +
geom_smooth(method = "lm", se = FALSE, color = "red") +
labs(
title = paste0("Mean Syuzhet Sentiment Over Time by - ", uni),
y = "Mean Sentiment Score",
x = "Month"
))
# Did not found a way to get the sentiment from the tweets for each language so I will use the full_text_emojis column and detect the language of the words only in german
# Tokenize and Preprocess Words
uni_words_en <- uni_tweets %>%
unnest_tokens(word, full_text_emojis) %>%
anti_join(get_stopwords(language = "en"), by = "word") %>%
distinct() %>%
filter(nchar(word) > 3) %>%
filter(!str_detect(word, "\\d")) %>%
filter(!str_detect(word, "https?://\\S+|www\\.\\S+|t\\.co|http|https"))
sentiment_words_en <- uni_words_en %>%
mutate(
sentiment = get_sentiment(word, method = "syuzhet")
)
# Separate Positive and Negative Words
positive_words_en <- sentiment_words_en %>%
filter(sentiment >= 0) %>%
count(word, sort = TRUE) %>%
rename(freq = n)
negative_words_en <- sentiment_words_en %>%
filter(sentiment < 0) %>%
count(word, sort = TRUE) %>%
rename(freq = n)
# Create and Display Word Clouds
# positive words
print(paste0("Positive words for: ", uni))
print(head(positive_words_en, 20))
print(wordcloud2(data.frame(
word = positive_words_en$word,
freq = positive_words_en$freq
), size = 0.5))
print(paste0("Negative words for: ", uni))
print(head(negative_words_en, 20))
# negative words
print(wordcloud2(data.frame(
word = negative_words_en$word,
freq = negative_words_en$freq
), size = 0.5))
}
## [1] "Positive words for: FHNW"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 FHNW aacsb 1
## 2 FHNW academy 1
## 3 FHNW acbtxijqdl 1
## 4 FHNW accelerate 1
## 5 FHNW accross 1
## 6 FHNW activities 1
## 7 FHNW adorno 1
## 8 FHNW agreement 1
## 9 FHNW aline 1
## 10 FHNW although 1
## 11 FHNW ambassador 1
## 12 FHNW america 1
## 13 FHNW american 1
## 14 FHNW among 1
## 15 FHNW amxxbmlfyc 1
## 16 FHNW anlass 1
## 17 FHNW anmelden 1
## 18 FHNW announce 1
## 19 FHNW announcement 1
## 20 FHNW announces 1
## [1] "Negative words for: FHNW"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 FHNW avoid 1
## 2 FHNW bacteria 1
## 3 FHNW blatant 1
## 4 FHNW blind 1
## 5 FHNW boom 1
## 6 FHNW breaking 1
## 7 FHNW challenge 1
## 8 FHNW cloud 1
## 9 FHNW competition 1
## 10 FHNW concerned 1
## 11 FHNW devastating 1
## 12 FHNW disadvantaged 1
## 13 FHNW exhausted 1
## 14 FHNW forget 1
## 15 FHNW hype 1
## 16 FHNW late 1
## 17 FHNW launch 1
## 18 FHNW limited 1
## 19 FHNW mistakes 1
## 20 FHNW outreach 1
## [1] "Positive words for: FH_Graubuenden"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 FH_Graubuenden able 1
## 2 FH_Graubuenden abroad 1
## 3 FH_Graubuenden abstract 1
## 4 FH_Graubuenden abstracts 1
## 5 FH_Graubuenden academic 1
## 6 FH_Graubuenden accu_rate 1
## 7 FH_Graubuenden across 1
## 8 FH_Graubuenden action 1
## 9 FH_Graubuenden activities 1
## 10 FH_Graubuenden address 1
## 11 FH_Graubuenden administration 1
## 12 FH_Graubuenden advantage 1
## 13 FH_Graubuenden advantages 1
## 14 FH_Graubuenden adventurous 1
## 15 FH_Graubuenden alliance 1
## 16 FH_Graubuenden almost 1
## 17 FH_Graubuenden alps 1
## 18 FH_Graubuenden already 1
## 19 FH_Graubuenden also 1
## 20 FH_Graubuenden alumnus 1
## [1] "Negative words for: FH_Graubuenden"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 FH_Graubuenden collective 1
## 2 FH_Graubuenden competition 1
## 3 FH_Graubuenden corruption 1
## 4 FH_Graubuenden countdown 1
## 5 FH_Graubuenden fall 1
## 6 FH_Graubuenden fallen 1
## 7 FH_Graubuenden fighting 1
## 8 FH_Graubuenden fire 1
## 9 FH_Graubuenden kick 1
## 10 FH_Graubuenden leave 1
## 11 FH_Graubuenden neglect 1
## 12 FH_Graubuenden neglecting 1
## 13 FH_Graubuenden problem 1
## 14 FH_Graubuenden quiz 1
## 15 FH_Graubuenden rainy 1
## 16 FH_Graubuenden risks 1
## 17 FH_Graubuenden spent 1
## 18 FH_Graubuenden strange 1
## 19 FH_Graubuenden stupidest 1
## 20 FH_Graubuenden sues 1
## [1] "Positive words for: ZHAW"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 ZHAW _bengraziano 1
## 2 ZHAW abilities 1
## 3 ZHAW able 1
## 4 ZHAW abroad 1
## 5 ZHAW abstracts 1
## 6 ZHAW academ 1
## 7 ZHAW academic 1
## 8 ZHAW acertainpain 1
## 9 ZHAW achievement 1
## 10 ZHAW across 1
## 11 ZHAW actually 1
## 12 ZHAW addition 1
## 13 ZHAW additional 1
## 14 ZHAW adespydvzf 1
## 15 ZHAW admits 1
## 16 ZHAW adopted 1
## 17 ZHAW advise 1
## 18 ZHAW advisory 1
## 19 ZHAW afterwards 1
## 20 ZHAW agarwaledu 1
## [1] "Negative words for: ZHAW"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 ZHAW barrier 1
## 2 ZHAW bastion 1
## 3 ZHAW break 1
## 4 ZHAW challenge 1
## 5 ZHAW cold 1
## 6 ZHAW competition 1
## 7 ZHAW conflict 1
## 8 ZHAW countdown 1
## 9 ZHAW desert 1
## 10 ZHAW economic 1
## 11 ZHAW enough 1
## 12 ZHAW entitled 1
## 13 ZHAW fled 1
## 14 ZHAW foreign 1
## 15 ZHAW hack 1
## 16 ZHAW hazard 1
## 17 ZHAW hidden 1
## 18 ZHAW ironic 1
## 19 ZHAW missing 1
## 20 ZHAW moan 1
## [1] "Positive words for: bfh"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 bfh _mich_i 1
## 2 bfh abend 1
## 3 bfh able 1
## 4 bfh academic 1
## 5 bfh accepted 1
## 6 bfh access 1
## 7 bfh across 1
## 8 bfh activities 1
## 9 bfh addressing 1
## 10 bfh administration 1
## 11 bfh agriculture 1
## 12 bfh alphasolarpro 1
## 13 bfh alternative 1
## 14 bfh always 1
## 15 bfh amarenabrown 1
## 16 bfh america's 1
## 17 bfh analysis 1
## 18 bfh andreasnaef 1
## 19 bfh annetwh 1
## 20 bfh announce 1
## [1] "Negative words for: bfh"
## # A tibble: 12 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 bfh aggression 1
## 2 bfh broken 1
## 3 bfh competition 1
## 4 bfh discrimination 1
## 5 bfh fail 1
## 6 bfh forget 1
## 7 bfh inequality 1
## 8 bfh labor 1
## 9 bfh player 1
## 10 bfh sorry 1
## 11 bfh stole 1
## 12 bfh stop 1
## [1] "Positive words for: hes_so"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 hes_so academic 1
## 2 hes_so active 1
## 3 hes_so actors 1
## 4 hes_so actualitesvd 1
## 5 hes_so administration 1
## 6 hes_so administrative 1
## 7 hes_so admission 1
## 8 hes_so advanced 1
## 9 hes_so agencies 1
## 10 hes_so agenda 1
## 11 hes_so also 1
## 12 hes_so amazing 1
## 13 hes_so amman 1
## 14 hes_so analyses 1
## 15 hes_so anne_ramelet 1
## 16 hes_so announce 1
## 17 hes_so anounce 1
## 18 hes_so antonio 1
## 19 hes_so applications 1
## 20 hes_so apply 1
## [1] "Negative words for: hes_so"
## # A tibble: 9 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 hes_so confessions 1
## 2 hes_so converted 1
## 3 hes_so fade 1
## 4 hes_so failure 1
## 5 hes_so hard 1
## 6 hes_so intense 1
## 7 hes_so launch 1
## 8 hes_so poor 1
## 9 hes_so vice 1
## [1] "Positive words for: hslu"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 hslu aacsb 1
## 2 hslu able 1
## 3 hslu abstract 1
## 4 hslu academia 1
## 5 hslu academic 1
## 6 hslu accept 1
## 7 hslu acceptance 1
## 8 hslu access 1
## 9 hslu according 1
## 10 hslu account 1
## 11 hslu accreditation 1
## 12 hslu achieving 1
## 13 hslu action 1
## 14 hslu additions 1
## 15 hslu address 1
## 16 hslu advantage 1
## 17 hslu afternoon 1
## 18 hslu agflow 1
## 19 hslu ahead 1
## 20 hslu aims 1
## [1] "Negative words for: hslu"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 hslu addiction 1
## 2 hslu awaited 1
## 3 hslu bacteria 1
## 4 hslu challenge 1
## 5 hslu competition 1
## 6 hslu crashes 1
## 7 hslu cut_up_tv 1
## 8 hslu dark 1
## 9 hslu dizzy 1
## 10 hslu error 1
## 11 hslu fail 1
## 12 hslu fall 1
## 13 hslu fears 1
## 14 hslu fire 1
## 15 hslu hack 1
## 16 hslu laden 1
## 17 hslu launch 1
## 18 hslu missed 1
## 19 hslu mistake 1
## 20 hslu regulatory 1
## [1] "Positive words for: ost_fh"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 ost_fh announce 1
## 2 ost_fh august 1
## 3 ost_fh backbone 1
## 4 ost_fh based 1
## 5 ost_fh beautiful 1
## 6 ost_fh bridge 1
## 7 ost_fh business 1
## 8 ost_fh campus 1
## 9 ost_fh closer 1
## 10 ost_fh collaboration 1
## 11 ost_fh cooling 1
## 12 ost_fh curious 1
## 13 ost_fh cyber 1
## 14 ost_fh deadline 1
## 15 ost_fh december 1
## 16 ost_fh delighted 1
## 17 ost_fh easily 1
## 18 ost_fh eastern 1
## 19 ost_fh electricity 1
## 20 ost_fh emits 1
## [1] "Negative words for: ost_fh"
## # A tibble: 2 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 ost_fh dire 1
## 2 ost_fh fire 1
## [1] "Positive words for: supsi_ch"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 supsi_ch _dreicast 1
## 2 supsi_ch abbg 1
## 3 supsi_ch abbgroupnews 1
## 4 supsi_ch abroad 1
## 5 supsi_ch academ 1
## 6 supsi_ch academia 1
## 7 supsi_ch academic 1
## 8 supsi_ch academies_ch 1
## 9 supsi_ch access 1
## 10 supsi_ch accoding 1
## 11 supsi_ch achieve 1
## 12 supsi_ch action 1
## 13 supsi_ch activator 1
## 14 supsi_ch activities 1
## 15 supsi_ch address 1
## 16 supsi_ch administration 1
## 17 supsi_ch administrations 1
## 18 supsi_ch admooajcib 1
## 19 supsi_ch advanced 1
## 20 supsi_ch advancedstudies 1
## [1] "Negative words for: supsi_ch"
## # A tibble: 20 × 3
## # Groups: university [1]
## university word freq
## <chr> <chr> <int>
## 1 supsi_ch barrier 1
## 2 supsi_ch cloud 1
## 3 supsi_ch cold 1
## 4 supsi_ch collision 1
## 5 supsi_ch critical 1
## 6 supsi_ch danger 1
## 7 supsi_ch demand 1
## 8 supsi_ch demanded 1
## 9 supsi_ch distracts 1
## 10 supsi_ch drug 1
## 11 supsi_ch economic 1
## 12 supsi_ch eth_rat 1
## 13 supsi_ch fabrication 1
## 14 supsi_ch fire 1
## 15 supsi_ch foreign 1
## 16 supsi_ch forget 1
## 17 supsi_ch government 1
## 18 supsi_ch hard 1
## 19 supsi_ch intense 1
## 20 supsi_ch launch 1Conclusion:
The analysis indicates that Swiss Universities of Applied Sciences exhibit diverse tweeting patterns in terms of content, style, and emotions. Tweets often focus on academic achievements, projects, and institutional news, with varying emotional tones across different universities. Recognizing these patterns can help universities optimize their social media strategies to better engage with their audiences. ## Question 4: What specific advice can you give us as communication department of BFH based on your analysis? How can we integrate the analysis of tweets in our internal processes, can you think of any data products that would be of value for us?
The comprehensive analysis of BFH’s tweets reveals several insights that can be leveraged to enhance the communication strategy.
Language Analysis:
BFH predominantly tweets in German, with 2760 tweets in this language. This aligns with the linguistic preferences of their primary audience.
Emoji Analysis:
The analysis of emoji usage shows that certain emojis are frequently used, which can be leveraged to increase engagement. Popular emojis like 🎓 (graduation cap) and 🚀 (rocket) often signify academic achievements and dynamic growth, resonating well with the audience.
Summary key insights from the analysis
# Language Analysis
tweets %>%
filter(university == "bfh") %>%
count(lang) %>%
arrange(desc(n))## # A tibble: 17 × 3
## # Groups: university [1]
## university lang n
## <chr> <chr> <int>
## 1 bfh de 2760
## 2 bfh <NA> 212
## 3 bfh en 97
## 4 bfh lb 62
## 5 bfh fr 31
## 6 bfh fy 8
## 7 bfh no 6
## 8 bfh nl 3
## 9 bfh af 2
## 10 bfh cy 2
## 11 bfh da 2
## 12 bfh ht 2
## 13 bfh it 2
## 14 bfh ru-Latn 2
## 15 bfh es 1
## 16 bfh gd 1
## 17 bfh mt 1# Emoji Analysis
emoji_count <- tweets %>%
top_n_emojis(full_text)
emoji_count %>%
mutate(emoji_name = reorder(emoji_name, n)) %>%
ggplot(aes(n, emoji_name)) +
geom_col() +
labs(x = "Count", y = NULL, title = "Top 20 Emojis Used")
insights <- list(
"Most Active Hours" = hours_with_most_tweets_by_uni,
"Most Active Days" = days_with_most_tweets_by_uni,
"Content Analysis" = head(words_freqs_de),
"Sentiment Analysis" = head(tweets_filtered$sentiment)
)Recommendations:
Based on the analysis, the following recommendations can be made to enhance BFH’s communication strategy: 1. Optimize Tweet Release Times: Based on the analysis of tweet activity, the most active hours for BFH are typically in the morning. Focusing on releasing tweets during these peak hours can maximize engagement. Scheduling important announcements and updates during these times will likely yield better visibility and interaction. 2. Focus on Specific Days for Announcements: The analysis shows that Tuesday is the most active day for BFH tweets. Leveraging this day for critical updates and major announcements can ensure they reach a wider audience. Aligning content release schedules with these high-activity days can enhance communication effectiveness. Sentiment Analysis: Sentiment analysis indicates the emotional tone of the tweets, helping tailor content to resonate positively with the audience. By understanding which types of tweets generate positive reactions, the communication team can craft messages that are more likely to be well-received. This could involve highlighting student achievements, successful projects, and positive institutional news. 4. Implement Topic Modeling: Topic modeling reveals the key themes prevalent in the tweets. For BFH, topics often include academic projects, student updates, and digital initiatives. Aligning the communication strategy to emphasize these themes can enhance relevance and engagement. Regularly updating the communication team on trending topics can help keep the content aligned with audience interests.
Integrating Tweet Analysis into Internal Processes:
To fully leverage these insights, the BFH communication department can integrate tweet analysis into their regular workflow: 1. Real-Time Analytics Dashboard: Implement a dashboard that tracks tweet performance, including engagement metrics, sentiment scores, and topic trends. This allows for real-time adjustments to the communication strategy. 2. Scheduled Reports: Generate weekly or monthly reports summarizing key metrics and insights. This helps the team stay informed about what content is performing well and where improvements can be made. 3. Content Calendar: Develop a content calendar that aligns tweet releases with peak engagement times and days. Incorporate findings from sentiment and topic analyses to plan content that resonates with the audience. 4. Feedback Loop: Establish a feedback loop where the communication team reviews analytics data and adjusts the strategy accordingly. Regular team meetings to discuss these insights can foster a more data-driven approach to communication.
Potential Data Products:
To further enhance the communication strategy, BFH can consider developing data products that provide additional value: 1. Engagement Prediction Tool: A tool that predicts the best times to tweet based on historical data, optimizing tweet scheduling for maximum engagement. 2. Sentiment Analysis Bot: An automated system that analyzes the sentiment of drafts before they are posted, ensuring that the tone is appropriate and likely to generate positive reactions. 3. Trend Tracker: A feature that identifies emerging topics and trends in real-time, allowing the communication team to quickly adapt and incorporate relevant themes into their messaging.
By integrating these recommendations and tools, BFH can enhance its communication strategy, ensuring that its messages are timely, relevant, and engaging for its audience.